My usecase - so to say - is learning to use ES while trying to build a new log storage.
I am still struggling to get the mappings right, which leads to the need to reindex a lot.
I would now like to learn to use the built-in ILM feature to make a hot-cold setup,
where data moves to cold after 6 months. Unfortunately, in the current state of affairs,
it seems that I'll have to wait 6 months from last reindex to do that.
While I don't have an argument for doing reindexation as part of the final workflow,
I'd still like to plea for the feature, because it will remove a barrier to learning ES I am facing.
Not a bad idea. True, I can mostly prototype the thing with shorter intervals.
So you are correct, lack of this feature is not actually preventing me from learning ILM.
It is just preventing me from going the easy "set up and forget" route. I won't be able to actually benefit from ILM until my indices stabilise enough so they won't need reindexing any more. Until that I'll have to move my indices manually. And actually it is even worse, because I recently clone-renamed about 50% of my indices withing few hours. If I leave ILM enabled, it will try to move all of them at once.
You can use the index setting index.lifecycle.origination_date to set a custom "age" for the index, so that if you were to reindex, you can tell ILM that it is a certain age without waiting for that long.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.